Show HN: Orkia – a Rust runtime where AI agents can't bypass governance (github.com)

Orkia, a newly announced Rust-based runtime, aims to provide enterprise-level governance for AI agents. This innovative platform allows organizations to deploy LLM-based business agents with built-in features like policy enforcement, audit trails, and trust scoring, ensuring compliance-grade control over agent behavior. Unlike other runtimes that address safety superficially, Orkia's approach is structural and governed by default, meaning agents cannot execute without defined policies—a fail-closed mechanism. Its architecture includes 27 focused crates, ranging from governance orchestration to multi-strategy information retrieval, providing a robust framework for business automation. Significantly, Orkia's governance model is designed for adaptability and security, allowing agents to start with limited autonomy that can be adjusted based on their demonstrated behavior. The runtime’s pipeline tracks every action through a multi-stage governance procedure, maintaining audit records and sensitivity labels that impact agent trust scores. These scores play a crucial role in determining operational autonomy, with violations leading to a decrease in trust, effectively managing the risk of AI actions in sensitive environments. By facilitating secure containerized execution of tools, Orkia enhances both compliance and security for enterprises looking to harness AI effectively without compromising governance.